Evaluating and Trusting AI Outputs